Coronavirus - Modelling Aspects Only

[johnhemming]

We also will find in retrospect that a number of people with Covid-19 will have died in care homes or potentially at home. We don't know how many and we won't also necessarily know whether Covid-19 was the cause or not.

[bionichamster]

It seems counterintuitive, but as the government increases the number of daily tests (which is a good thing) the number of new positive cases will also go up (which looks like a bad thing, but actually it won't be, because there will be more new cases detected, not actually more new cases in reality). Hope this makes sense, happy to explain further if it doesn't.

It also means that it’s highly likely that the ratio of mild to serious cases being detected will shift in favour of the former.

Bh

[Itsallaguess]

It seems counterintuitive, but as the government increases the number of daily tests (which is a good thing) the number of new positive cases will also go up (which looks like a bad thing, but actually it won't be, because there will be more new cases detected, not actually more new cases in reality). Hope this makes sense, happy to explain further if it doesn't.

It also means that it’s highly likely that the ratio of mild to serious cases being detected will shift in favour of the former.

That's a good point too - that we might expect if these additional 'Daily new Cases' figures were from people with much more distressing cases of the virus, that they'd have already been in a hospital somewhere and appeared on the earlier test-results...

Cheers,

Itsallaguess

[zico]

Interesting new data from France. Twitter link below, but I've copied it out below for ease of reading.
(Note - most of what's below is for hospital data only. Adding nursing home Covid-19 deaths increases total deaths by 36%. Comparable figure for UK was 20%, but I can't find the report on the Office for National Statistics (ONS) website. ONS figures were based on small numbers, something like 210 v 190 so French %age is probably a much better estimate of what the UK %age will be, which means far more overall deaths.)

https://twitter.com/PedderSophie/status ... 4684690434

Short thread on France and covid-19. It increasingly looks as if France will not be following as awful a trajectory as Italy and Spain. This is showing up in a series of tentatively encouraging data points over the past 4-6 days

The increase in new hospital admissions for covid-19 in France has been slowing since April 1. This hints that nearly 3 weeks of confinement is starting to work:
Mar 31: +1749
April 1: +1882
April 2: +1607
April 3: +1186
April 4: +711

The rise in no. in intensive care has slowed for 5 consecutive days. The situation remains critical, but France’s hospitals are just about coping:
Mar 29: +359
Mar 30: +475
Mar 31: +458
April 1: +452
April 2: +382
April 3: +280
April 4: +176

Finally, no. of daily hospital deaths in France has been +/- 500 since March 30. No clear trend yet, but these numbers are expected to lag hospital admissions and no. in intensive care. (UK daily hospital deaths overtook France on April 1.

If French hospital covid-19 trends are tentatively encouraging, less so is the dire pattern unfolding in nationwide retirement homes . By April 4 these recorded 2,028 deaths, which aren’t included in France’s hospital death total (5,532)

All figures are from the French Ministry of Health, updated daily here https://dashboard.covid19.data.gouv.fr. Analysis of the advantages and limits of France’s centralised system in dealing with covid-19 in @TheEconomist
is free to read here https://economist.com/europe/2020/04/04 ... o-covid-19

[johnhemming]

This hints that nearly 3 weeks of confinement is starting to work:

The difficulty is that apart from Sweden which is not really a control experiment there is no control experiment.

Most importantly we have no idea of what proportion of the population have had the infection. If we are getting towards the 20-40% mark the infection rate will start slowing down.

Secondly it would take a lot of work to get into the figures of deaths. We know in the UK that quite a few people are missed out and also that the delay in reporting a death varies.

Hence "hints" is the best we can have at the moment.

[GoSeigen]

Most importantly we have no idea of what proportion of the population have had the infection. If we are getting towards the 20-40% mark the infection rate will start slowing down.

I don't think this is true. Cases resulting in death seem pretty clustered geographically, a sign of the effectiveness of lockdowns, contact tracing etc. If a large proportion of people had contracted the virus then clusters would be much more widespread than they are. So I think it's fairly clear that a small proportion has been infected. That's my amateur view anyway.

I'd say we're at the very early stages, have to see what happens in Asia esp India/Bangladesh/Indonesia, Africa and South America. Oh I know those people don't count but still...


GS

[dealtn]

I don't think this is true. Cases resulting in death seem pretty clustered geographically...

Where are you seeing the data for this conclusion please?

[johnhemming]

I don't think this is true.

There are different views on this.

https://www.ft.com/content/5ff6469a-6dd ... bea055720b

The new coronavirus may already have infected far more people in the UK than scientists had previously estimated — perhaps as much as half the population — according to modelling by researchers at the University of Oxford.

https://order-order.com/2020/04/01/20-t ... ronavirus/

New polling from Ipsos MORI has found that 20% of Britons think it is likely that they have already had COVID-19. 6% are confident about having had the disease, whilst 14% think it is fairly likely.

I think I saw something from Imperial talking about 2%.

I think the opinion poll is quite significant as it is what people think has happened to themselves.

I don't think a case has been made that we are doing anything other than reducing the speed of infection. It is still most likely that 50%+ of people will get it and we need to ensure those most at risk are protected from harm.

I have found the imperial report at 2.7%
https://www.telegraph.co.uk/global-heal ... es-europe/

[zico]

I've updated my model for hospitalised Covid-19 deaths (excluding nursing home or other Covid-19 deaths) assuming a 7-day lag between %increase in "new cases" and "new deaths", and using the percentage decline in "new cases" from Italy, assuming the UK has reached its peak daily increase in new cases at 4,400 (so essentially ignoring yesterday's 5,900 increase).
The results look significantly worse than the reported figures for Italy and Spain, so I really hope this means my model is overestimating deaths.

Both Italy and Spain are showing fairly sharp declines in daily deaths after reaching their peaks, whereas my model shows a slower reduction.
(I suspect this may be because an increasing % of deaths are occurring in care homes, or not being recorded as Covid-19 deaths, but again, I hope I am wrong in this).

Italy and Spain have peak daily deaths of around 950, with sharp falls afterwards.
My model predicts a flatter peak for the UK with around 1,100 deaths at peak, followed by a gradual decline to 700 daily deaths by April 20th (which is obviously still massively high). For the next 2 days, I estimate UK deaths to be 658, then 759, but then increasing sharply to 1,000+ by Wednesday.

On a positive note, Spain & Italy experience is probably a much better guide for what will happen in the UK than my little model, because they are all western democracies who have introduced some kind of lockdown, and managed to get beyond their peak new cases and peak deaths.
Although Spain only has 40 million population, and Italy 60 million population, the total population size is probably not that important at this stage of the spread of the disease.

(If anyone can let me know how to use the "Img" tag, I'd be grateful. I tried putting my Imgur graph into it, but just got the word "Image".)

[Itsallaguess]

(If anyone can let me know how to use the "Img" tag, I'd be grateful. I tried putting my Imgur graph into it, but just got the word "Image".)

I PM'd you some IMGUR/tag instructions on Saturday 4th - did you give them a go?

If there's anything unclear, just pretend you're going to reply to the PM using the quote button, and examine the underlying post-edit window, and it should all then become clear...

Let me know how you get on..

Cheers,

Itsallaguess

[dspp]

(If anyone can let me know how to use the "Img" tag, I'd be grateful. I tried putting my Imgur graph into it, but just got the word "Image".)

Zico,

Load your image on imgur.

Write a post here in TLF.

Use the Img button in the TLF text editor to insert the img start and end codes. Between the codes paste your URL from imgur.

Use the correct imgur link. Use the second one down, the "Direct Link".

If you do it right the actual image will display here in your post on TLF.

regards, dspp

[zico]

Itsallaguess, many thanks for your help in displaying the graph. I've now attached the graph I referred to in my previous post.
Note that after 11th April, the estimates are very straight line, and that's because I assumed a constant 2% drop every day on the previous highest "new cases" total of 4,300. The estimates before 11th April use the actual %increases in new cases from 7 days earlier.

[zico]

Latest daily figures
13,069 tests
3,802 positive
439 deaths.

Looks like yesterday's 5,903 positive cases was a huge outlier. Someone said earlier on this thread that probably more NHS workers with mild symptoms are being tested.
% of positive tests down to 29% from 40%+ in last 3 days.

Surprising (but welcome) reduction in deaths figure, though probably mainly due to weekend recording dip (as happened last week) - deaths aren't included in total before next-of-kin have given permission for them to be recorded as such, so may be some delay in this.

[Sorcery]

Latest daily figures
13,069 tests
3,802 positive
439 deaths.

Looks like yesterday's 5,903 positive cases was a huge outlier. Someone said earlier on this thread that probably more NHS workers with mild symptoms are being tested.
% of positive tests down to 29% from 40%+ in last 3 days.

Surprising (but welcome) reduction in deaths figure, though probably mainly due to weekend recording dip (as happened last week) - deaths aren't included in total before next-of-kin have given permission for them to be recorded as such, so may be some delay in this.

Are you sure about next of kin permission being required for including in the numbers?
I could understand the requirement for giving out the names but a number?
It would be grossly unscientific to have the numbers distorted by next of kin refusal.

[zico]

Are you sure about next of kin permission being required for including in the numbers?
I could understand the requirement for giving out the names but a number?
It would be grossly unscientific to have the numbers distorted by next of kin refusal.

Yes, I'm sure that's what the government is doing.
I entirely agree with your point.

[Sorcery]

Are you sure about next of kin permission being required for including in the numbers?
I could understand the requirement for giving out the names but a number?
It would be grossly unscientific to have the numbers distorted by next of kin refusal.

Yes, I'm sure that's what the government is doing.
I entirely agree with your point.

Have done a search for "uk next of kin permission for inclusion in coronovirus numbers" without finding anything relevant. Do you have a link?
Think there is a lag for deaths outside hospital inclusion, but what you are saying would be outrageous and likely to lead to incorrect decision making.

[ChrisNix]

Itsallaguess, many thanks for your help in displaying the graph. I've now attached the graph I referred to in my previous post.
Note that after 11th April, the estimates are very straight line, and that's because I assumed a constant 2% drop every day on the previous highest "new cases" total of 4,300. The estimates before 11th April use the actual %increases in new cases from 7 days earlier.

This would be a similar shape to Italy. Will be interesting to see if today's figure is weekend aberrant.

After allowing for lags and population differences, UK (actual), French, Italian and Dutch figures around the same level as of tonight.

[zico]

Are you sure about next of kin permission being required for including in the numbers?
I could understand the requirement for giving out the names but a number?
It would be grossly unscientific to have the numbers distorted by next of kin refusal.

Yes, I'm sure that's what the government is doing.
I entirely agree with your point.

Have done a search for "uk next of kin permission for inclusion in coronovirus numbers" without finding anything relevant. Do you have a link?
Think there is a lag for deaths outside hospital inclusion, but what you are saying would be outrageous and likely to lead to incorrect decision making.

Here's a link to a Newsnight video explaining the change.

https://twitter.com/siennamarla/status/ ... 0202818561

I'm sure the government scientists will have correct up-to-date figures for their decision making - which is what's important.
I would expect that every country in the world will be looking to put some kind of a gloss on their public figures. For example, both Italy and Spain had a particularly high peak daily deaths which was followed next day by a dramatic reduction.
The methodology for the published UK daily deaths just means that the figures are "lumpy" rather than following an expected statistical curve.
The published figures also exclude non-hospitalisation Covid-19 deaths, which would add 20% to UK figures (Source:ONS weekly report) or 35% (using France figures for deaths outside hospitals).

[Sorcery]

Yes, I'm sure that's what the government is doing.
I entirely agree with your point.

Have done a search for "uk next of kin permission for inclusion in coronovirus numbers" without finding anything relevant. Do you have a link?
Think there is a lag for deaths outside hospital inclusion, but what you are saying would be outrageous and likely to lead to incorrect decision making.

Here's a link to a Newsnight video explaining the change.

https://twitter.com/siennamarla/status/ ... 0202818561

I'm sure the government scientists will have correct up-to-date figures for their decision making - which is what's important.
I would expect that every country in the world will be looking to put some kind of a gloss on their public figures. For example, both Italy and Spain had a particularly high peak daily deaths which was followed next day by a dramatic reduction.
The methodology for the published UK daily deaths just means that the figures are "lumpy" rather than following an expected statistical curve.
The published figures also exclude non-hospitalisation Covid-19 deaths, which would add 20% to UK figures (Source:ONS weekly report) or 35% (using France figures for deaths outside hospitals).

The twitter link does not work for me, a page not found error. The "next of kin" issue would make a good question for politicians in the daily briefings if a reporter was present that had the wit to ask it. I suppose we know reporters are not good on science though.

[zico]

This link should work. Sorry, didn't check the previous one worked before I posted it.

https://twitter.com/siennamarla/status/ ... 0202818561